Image Super-resolution Reconstruction Based on Sparse Representation and Low-rank Matrix Completion

摘要

This paper addresses the problem of generating a High-resolution (HR) image from a single Low-resolution (LR) image. We propose the super-resolution reconstruction approach based on sparse representation and low-rank matrix completion. The approach represents images in forms of sparse and rearranges image regions into the low dimension construction-matrices of low rank. High-frequency details of image are restored using the sparse representation which is recovered from the down-sampled images. For paths at the same position of multiple pictures which are obtained by several super resolution reconstructions are highly correlated, they are arranged to be a matrix of low-rank which can be completed exactly from corrupted entries. Experiment results demonstrate that the proposed method significantly improves the PSNR and visual quality of reconstructed high-resolution images.